meta-nao - Yocto layer for Softbank Nao V6

This layer provides support for Softbank Nao V6 robot to construct a custom image as an alternative to the original image in context of RoboCup SPL.

In general this image should be usable at this point, and if it isn't this is considered a bug and we appreciate a report by opening an issue. As this is a very first attempt with the Yocto project and image construction for the Nao robot, Team HULKs is also very interested in suggestions regarding code quality and style as well as best-practices.

Getting Started

1. Create a directory for the upcoming setup and build phases e.g. worktree/. Make sure you have at least 100GB empty disk space available!

2. For project setup the siemens/kas framework is used. To setup kas use the containerized (podman/docker) version via the kas-container script provided here and store it inside the worktree directory. Alternatively setup kas via a python-pip installation, follow the installation steps in the user guide.

3. meta-nao ships multiple kas-project.yml project description file. This file defines the project structure kas has to setup for the Yocto build phase. Clone the meta-nao repository into some directory used for the Yocto build e.g. worktree/meta-nao.

git clone [email protected]:hulks/meta-nao worktree/meta-nao

4. The Nao V6 uses LoLA and HAL for the communication with the chestboard. All these binaries and libraries necessary to operate the Nao properly are shipped with the .opn robocupper image and not included in this repository. To acquire the necessary binaries the meta-nao/recipes-support/aldebaran/extract_binaries.sh script is used. This script fetches all binaries from inside the robocupper image and collects them in an archive for the upcoming build phase. To generate the archive containing the aldebaran binaries run:

cd meta/recipes-support/aldebaran/
mkdir -p aldebaran-binaries
./extract_binaries.sh -o aldebaran-binaries/aldebaran_binaries.tar.gz nao-x86-firmware-249_20190503_203829_robocupper.opn

The script references the original robocupper image shipped by softbank. Contact the RoboCup SPL TC to get this image. If you get errors regarding libguestfs and supermin, try running ./extract_binaries.sh with root permissions again.

5. Execute kas from inside the worktree directory referencing the project description files to enter the build environment

./kas-container shell meta-nao/kas/base.yml:meta-nao/kas/hulks.yml

kas fetches all necessary build sources and sets them up in the respective worktree directory. After kas has setup the working directory, your directory structure should look like this:

worktree/
├── build
├── meta-clang
├── meta-congatec-x86
├── meta-intel
├── meta-nao
├── meta-openembedded
└── poky

If your container shell does not work as expected, you might have to set your TERM environment variable properly (e.g. TERM=xterm).

6. Build a nao image. The Yocto project uses BitBake for task execution. Call the following build command from inside the kas container shell:

bitbake nao-image

This generates and executes all necessary tasks and targets to construct a proper .opn file. This build phase might take several hours depending on the performance of your build machine and your internet connection. BitBake uses a very elaborated caching strategy to speed up following builds of targets. Thus small changes afterwards can only take a few minutes.

7. Fetch and deploy the image. After BitBake ran all tasks up to nao-opn a new .opn file is generated in worktree/build/tmp/deploy/images/nao-v6/nao-image-HULKs-OS-[VERSION].ext3.gz.opn. To setup a flash stick run:

dd if=image_path.opn of=/dev/sdb bs=4M status=progress oflag=sync

Questions and answers

How can I log into the robot?

The login credentials for user root and nao are defined in meta-nao/meta/recipes-core/images/nao-image.bb. Per default the nao user has an empty password. The password of the root user is root.

How can I configure IP addresses?

Networking is controlled via systemd-networkd and respective .network units. You can configure the network via wpa_supplicant.conf and respective .network files. Have a look at meta-nao/meta/recipes-conf/nao-wifi-conf/nao-wifi-conf/80-wlan.network and meta-nao/meta/recipes-conf/nao-wifi-conf/nao-wifi-conf/wpa_supplicant-nl80211-wlan0.conf.

How can I customize the image?

The Yocto Project is organized in layers. You can edit the existing meta-nao/meta layer or (better) add an additional layer alongside meta-nao, meta-openembedded, etc. to overlay configuration in other layers. Have a look at meta-example for an example overlay extending the nao image by the boost library. You can use https://layers.openembedded.org/layerindex/branch/master/recipes/ to search for existing recipes.

How do I target the nao architecture during development?

The Yocto project contains tasks to build a proper SDK to use for development. To build the SDK do the following:

1. Enter the build container

./kas-container -d shell meta-nao/kas/base.yml:meta-nao/kas/hulks.yml

You can append the aarch64.yml project description to configure Yocto for building a aarch64 SDK.

2. Build the SDK (from inside the container)

bitbake -c populate_sdk nao-image

This again takes several hours. On machines at HULKs this can take up to 4 hours.

3. Download and install the SDK

After a successful build, the SDK is located at worktree/build/tmp/deploy/sdk/HULKs-OS-[MACHINE]-toolchain-[VERSION].sh. To install the SDK run the script and follow the instructions. Afterwards you are able to source the build environment and use the respective cross compilers.

Links and Resources

Yocto Project Mega-Manual

kas user guide

Open Embedded recipe search